Path: utzoo!attcan!uunet!tank!ncar!mailrus!cornell!batcomputer!gdykes
From: gdykes@batcomputer.tn.cornell.edu (Gene Dykes)
Newsgroups: comp.windows.x
Subject: Xtk "unsigned int" typedefs considered harmful
Message-ID: <6289@batcomputer.tn.cornell.edu>
Date: 12 Sep 88 18:54:12 GMT
Distribution: comp
Organization: Theory Center, Cornell U., Ithaca NY
Lines: 43

The Facts
---------
"Dimension", "Position", and "Cardinal" are typedef'd as unsigned ints
in lib/Xt/Intrinsics.h.

The Problem
-----------
The following code fragment is a very reasonable thing to find in a geometry
manager widget:

	Dimension new_width, old_width, adjustment ;
	Cardinal  number_of_children ;
	...
	new_width = old_width + adjustment / number_of_children ;

I think that it's reasonable to think of an "adjustment" to a Dimension
as being of type Dimension, too.  But if adjustment is, say, -100, then
this code clearly won't work.  But,  for the sake of argument, let's say
that adjustment should definitely be an "int".  Then, the code above still
croaks out, and I think that this violates the principle of least surprise.
One needs a cast as follows:

	new_width = old_width + adjustment / (int) number_of_children ;

I got tired of trying to find every place that needed a cast in my widgets,
so I threw up my hands and put the following in my header file:

    #define Dimension int
    #define Position  int
    #define Cardinal  int

There are probably other circumstances like this, and I contend that there will
soon be lots of these bugs lurking around, just waiting for the unexpected
negative number to come along and blow the program away.

Isn't it worth limiting Dimensions to two billion or so, by specifying them
as "int" instead of "unsigned int"? I would presume that "Cardinal", etc.,
are typedef'd because the programmer is not supposed to need to know
what they really are, when in fact that ignorance leads to big trouble.

(My apologies if this has been hashed out before...)
-- 
Gene Dykes, gdykes@tcgould.tn.cornell.edu